Research Engineer, Forge
Lausanne · Künstliche Intelligenz · Technologieunternehmen · Englisch
About Mistral
Mistral provides full-stack AI solutions: from frontier models to developer tools, applications, and compute. We partner with enterprises tackling the hardest problems across high-stakes industries like finance, manufacturing, defense, healthcare, and the public sector, co-creating customized AI systems that they can run on their terms. We are a dynamic, collaborative team passionate about AI and its potential to transform society. Our diverse workforce thrives in competitive environments and is committed to driving innovation. Our teams are distributed between Europe, North America, Asia and the Middle East. We are creative, low-ego and team-spirited.
What you will do
Build and improve post‑training and evaluation workflows (CPT/SFT/RL/distillation), turning prototypes into repeatable Forge “recipes”. Develop tools and pipelines for synthetic data generation, data curation, training, evaluation, and deployment. Debug and harden large‑scale ML systems: distributed training, scheduling/execution, checkpointing, observability, and reproducibility. Improve the Forge codebase via clear APIs, tests, documentation, and maintainable abstractions. Push the frontier of our RL training stack (e.g., high-throughput async rollout and scalable post‑training systems at frontier-model scale) Make sure Forge deployment is seamless and adaptable to a diversity of clients (hardware access, software stack, cloud and on-premises, …) Partner with researchers and infrastructure engineers to translate bottlenecks into concrete system improvements.
About you
Strong Python engineering skills and experience working in large codebases (testing, code review, CI, operational ownership). Hands‑on experience with PyTorch, JAX, or similar. Strong systems and infrastructure fundamentals. Experience with LLM training or post‑training: fine‑tuning, RL, distillation, evaluation, and/or data pipelines. Excellent debugging skills in ambiguous systems (distributed jobs, data issues, quality regressions, infra failures). Clear communication with technical and non‑technical stakeholders. High agency, low ego, and comfort in fast‑moving, under‑specified environments. Distributed training experience (FSDP, DeepSpeed, Megatron, etc.). Cluster/orchestration experience (SLURM, Ray, Kubernetes, Kueue, Karpenter, Skypilot, etc.). Experience building reliable ML infrastructure, evaluation systems, or large‑scale data processing pipelines. Research experience in LLMs, agents, multimodal models, reasoning, code, or domain adaptation. Open‑source contributions, publications, or widely used internal tooling. Experience training multi‑billion‑parameter models (pre‑training or RL). Experience training on petabyte- and exabyte-scale datasets. Ability to identify bottlenecks across the stack and drive improvements from first principles.
Fähigkeiten
Gehaltsspanne
Auf AnfrageDieser Arbeitgeber nennt die Gehaltsspanne im ersten Gespräch.
